Improve `fit_class_random_forest` and `fit_regr_random_forest` support

[soxofaan]

A couple of issues with current (experimental) implementation of fit_class_random_forest and fit_regr_random_forest

• [x] Add fit_regr_random_forest
• [ ] python client defines default max_variables=null , while that's not the case in current process specs, see https://github.com/Open-EO/openeo-processes/issues/365
• [x] methods fit_class_random_forest and fit_regr_random_forest are defined on (raster) DataCube object, while it should be on vector cube. Underlying problem is that aggregate_spatial returns a (raster) DataCube instead of VectorCube

[soxofaan]

methods fit_class_random_forest and fit_regr_random_forest are defined on (raster) DataCube object, while it should be on vector cube. Underlying problem is that aggregate_spatial returns a (raster) DataCube instead of VectorCube

handled by

• 386